Introduce a new AccessorPair type for handling JavaScript accessors.

Introduce a new AccessorPair type for handling JavaScript accessors.

Using a seaparate type for this instead of a 2-element FixedArray has some
advantages: First of all AccessorPair is one word smaller, which is not much,
but anyway. The real advantage of this change is that we get a better typing
internally, making it much easier to find and modify all the places having to do
with JavaScript accessors.

This change is part of a series introducing map sharing for objects with
JavaScript accessors. More to come...

[Kevin Millikin (Chromium), 2012-01-10 08:57:29]

LGTM.  My concern is that the names AccessorPair and AccessorInfo are too
similar and it's not obvious why the difference.  We should probably try to come
up with a better name for AccessorInfo---they are the accessors defined through
the API, and the getter/setter are C++ functions.  We might also change the type
of  getter and setter in AccessorInfo from Object to Foreign to make it more
obvious.

http://codereview.chromium.org/9152001/diff/1001/src/factory.h
File src/factory.h (right):

http://codereview.chromium.org/9152001/diff/1001/src/factory.h#newcode72
src/factory.h:72: Handle<AccessorPair> NewAccessorPair();
Maybe a comment here and also in heap.h about why these are always pretenured
(I'm not sure why they are).

I'm also not sure why we pretenure all structs.  It's not obviously best.

http://codereview.chromium.org/9152001/diff/1001/src/heap.h
File src/heap.h (right):

http://codereview.chromium.org/9152001/diff/1001/src/heap.h#newcode619
src/heap.h:619: MaybeObject* AllocateAccessorPair();
MUST_USE_RESULT to warn if a developer does not check for allocation failure.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4611
src/objects.cc:4611: if (!maybe_accessors->To<Object>(&accessors)) return
maybe_accessors;
Huh.  Maybe we should get rid of MaybeObject::ToObject since it is the same
implementation (minus the Object::cast) as MaybeObject::To<Object>.  Or else
continue using ToObject.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4736
src/objects.cc:4736: AccessorPair *accessors = AccessorPair::cast(element);
Misplaced *.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4754
src/objects.cc:4754: AccessorPair *accessors = AccessorPair::cast(obj);
Misplaced *.

[Sven Panne, 2012-01-10 09:45:03]

http://codereview.chromium.org/9152001/diff/1001/src/factory.h
File src/factory.h (right):

http://codereview.chromium.org/9152001/diff/1001/src/factory.h#newcode72
src/factory.h:72: Handle<AccessorPair> NewAccessorPair();
On 2012/01/10 08:57:30, Kevin Millikin wrote:
> Maybe a comment here and also in heap.h about why these are always pretenured
> (I'm not sure why they are).
> 
> I'm also not sure why we pretenure all structs.  It's not obviously best.

Done.

http://codereview.chromium.org/9152001/diff/1001/src/heap.h
File src/heap.h (right):

http://codereview.chromium.org/9152001/diff/1001/src/heap.h#newcode619
src/heap.h:619: MaybeObject* AllocateAccessorPair();
On 2012/01/10 08:57:30, Kevin Millikin wrote:
> MUST_USE_RESULT to warn if a developer does not check for allocation failure.

Done.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc
File src/objects.cc (right):

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4611
src/objects.cc:4611: if (!maybe_accessors->To<Object>(&accessors)) return
maybe_accessors;
On 2012/01/10 08:57:30, Kevin Millikin wrote:
> Huh.  Maybe we should get rid of MaybeObject::ToObject since it is the same
> implementation (minus the Object::cast) as MaybeObject::To<Object>.  Or else
> continue using ToObject.

IMHO ToObject should be on death row... >:-) The templated version is much more
useful, so we can remove ToObject in a separate CL.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4736
src/objects.cc:4736: AccessorPair *accessors = AccessorPair::cast(element);
On 2012/01/10 08:57:30, Kevin Millikin wrote:
> Misplaced *.

Done.

http://codereview.chromium.org/9152001/diff/1001/src/objects.cc#newcode4754
src/objects.cc:4754: AccessorPair *accessors = AccessorPair::cast(obj);
On 2012/01/10 08:57:30, Kevin Millikin wrote:
> Misplaced *.

Done.